Speech synthesis (TTS: Text-to-Speech) generates artificial voice from text input. In telephony, it powers IVR voice guidance, voicemail transcription readback, and screen readers for visually impaired users.
The technology has evolved rapidly. Early synthetic voices sounded mechanical, but deep learning models (WaveNet, VALL-E) now produce speech nearly indistinguishable from humans. Emotional expression, natural intonation, and individual voice reproduction enable narration and customer support applications.
Misuse is a serious concern. Voice clone scams generate convincing synthetic voices from seconds of audio samples to impersonate family members demanding money. Combined with voice recognition, real-time synthetic conversations are becoming possible.
Defenses include family code words, hanging up and calling back when money is mentioned, and using one-time passwords alongside voice authentication. See AI voice clone scams for current tactics.